A method and system for predicting unbalanced sample galloping based on meteorological data

Abstract

This invention relates to the field of power transmission engineering technology and discloses a method and system for predicting unbalanced galloping events based on meteorological data. The prediction method includes: collecting historical monitoring data from each monitoring point and simultaneously collecting meteorological information from the corresponding monitoring points; selecting one sample data from the daily data of each device, using the monitoring device and date as the basic unit; grouping the non-galloping samples in the sample data according to the line number, and retaining only samples corresponding to different spans within each group; assigning differentiated loss weights to the sample data based on the span characteristics, high-incidence periods of galloping, and typical meteorological combinations, and constructing a weighted cross-entropy loss function; constructing a daily-scale galloping prediction model, and training the model using the weighted cross-entropy loss function to achieve daily-scale prediction of galloping events. By constructing a scientific sample extraction strategy, introducing a category weight mechanism, and adopting a deep learning classification model, the model's ability to identify minority events is improved.

Description

Technical Field

[0001] This invention relates to the field of power transmission engineering technology, specifically to a method and system for predicting unbalanced sample galloping based on meteorological data. Background Technology

[0002] Transmission line galloping refers to the large-amplitude, low-frequency mechanical oscillation of conductors under certain meteorological conditions, especially after icing or frost accumulation, triggered by crosswinds. Galloping typically takes the form of a pendulum, ellipse, or torsion, with frequencies as low as 0.1Hz to 3Hz and amplitudes reaching tens of centimeters or even several meters. Its occurrence mechanism is complex and closely related to various meteorological factors such as conductor tension, icing morphology, wind speed and direction, temperature, and humidity. Due to the strong nonlinear characteristics during galloping, it can trigger conductor jumping, phase-to-phase collisions, insulation breakdown, and even serious accidents such as conductor strand breakage and tower damage within a short period of time.

[0003] For a long time, conductor galloping has been a major hidden danger to the safe operation of power systems. Especially in the cold, icy regions of Northeast, Northwest, and North China, galloping occurs frequently in winter, putting immense pressure on power grid operation and emergency maintenance. To address this, power systems have widely deployed numerous online monitoring devices along transmission lines to monitor conductor displacement, acceleration, and galloping status in real time. Although the monitoring technology is relatively mature, because galloping events do not occur for most of the time, the number of galloping samples in the actual collected data is far less than the number of non-galloping samples, resulting in a typical extremely unbalanced data structure.

[0004] Faced with such a large-scale imbalanced data, traditional machine learning and statistical analysis methods often suffer from "biased majority class" during training, resulting in low accuracy and high false negative rate in dance category recognition, making it difficult to meet the accurate early warning requirements of engineering applications. Summary of the Invention

[0005] In view of this, the present invention provides a method and system for predicting unbalanced sample agitation based on meteorological data, in order to solve the problem that the agitation sample structure is extremely unbalanced, resulting in low accuracy of agitation category identification and high false negative rate.

[0006] In a first aspect, the present invention provides a method for predicting imbalanced sample agitation based on meteorological data, the method comprising:

[0007] Multiple monitoring devices are deployed along the transmission line to collect historical monitoring data from each monitoring point and simultaneously collect meteorological information from the corresponding monitoring points.

[0008] Using monitoring devices and dates as basic units, a sample data point is selected from the daily data of each device. The non-galloping samples in the sample data are grouped by line number. Only samples corresponding to different ranges are retained in each group. All galloping samples in the sample data are retained. The sample data contains complete time-series data for the day.

[0009] Based on the characteristics of the line span, the high-incidence period of galloping and typical weather combinations, the sample data are assigned differentiated loss weights, and a weighted cross-entropy loss function is constructed.

[0010] An LSTM network is used to construct a daily-scale gobling prediction model. A time-series input matrix is ​​constructed based on the selected sample data. The time-series input matrix is ​​input into the daily-scale gobling prediction model. The weighted cross-entropy loss function is used to train the daily-scale gobling prediction model to achieve daily-scale prediction of gobling events.

[0011] This invention provides an imbalanced sample gobling prediction method based on meteorological data. In view of the scarcity and spatiotemporal dispersion of gobling samples, a sample extraction strategy based on the monitoring device-time dimension is designed. Combined with class weight factors and deep learning models, the model's ability to identify minority events is improved.

[0012] In one optional implementation, differential loss weights are assigned to the sample data based on line span characteristics, high-incidence periods of galloping, and typical weather combinations, including:

[0013] Extract categorical features from the meteorological information, and calculate the category frequency weighting factor based on the frequency of each category in the sample data;

[0014] Based on the differences in the risk of galloping at different gear intervals, corresponding structural sensitivity factors are set for different gear intervals;

[0015] Based on the probability of the dance occurring in different time periods, corresponding time-sensitivity factors are set for different time periods;

[0016] For sample data that meet the preset meteorological combination, set meteorological condition adjustment factors.

[0017] In one optional implementation, the weighted cross-entropy loss function is as follows:

[0018]

[0019] in, The weighted cross-entropy loss function is... The final sample weights for multi-factor synthesis. For the true labels of the i,j,t,k samples, This represents the probability of a gobling event predicted by a daily-scale gobling prediction model.

[0020] In an optional implementation, the method further includes:

[0021] The numerical features in the meteorological information are normalized, including wind speed, temperature, humidity, and precipitation.

[0022] The categorical features in the meteorological information are transformed into uniquely coded features, which include wind direction category and precipitation category.

[0023] In one alternative implementation, all sample data are indexed using a combination of device number and timestamp.

[0024] In one optional implementation, the daily-scale gobling prediction model uses a one- or two-layer LSTM network to process the time-series input matrix, extracts time features, maps them to a single output node through a fully connected layer, and then outputs the probability of gobling occurring on that day through a Sigmoid activation function.

[0025] In one optional implementation, the diurnal agitation prediction model is expressed as follows:

[0026]

[0027] in, It is the Sigmoid activation function. This is the weight matrix of the fully connected layer. This represents the hidden state of the LSTM network at the last time step. This is a bias term.

[0028] Secondly, the present invention provides an imbalanced sample agitation prediction device based on meteorological data, the device comprising:

[0029] The data acquisition module is used to deploy multiple monitoring devices along the transmission line to collect historical monitoring data from each monitoring point and simultaneously collect meteorological information from the corresponding monitoring points.

[0030] The data filtering module is used to select one sample data from the daily data of each device, using the monitoring device and date as the basic unit. The non-vibrating samples in the sample data are grouped by line number, and only samples corresponding to different ranges are retained in each group. All vibrating samples in the sample data are retained. The sample data contains complete time-series data for the day.

[0031] The weight assignment module is used to assign differentiated loss weights to sample data based on line span characteristics, high-incidence periods of galloping, and typical weather combinations, and to construct a weighted cross-entropy loss function.

[0032] The model prediction module is used to construct a daily-scale gobling prediction model using an LSTM network. It constructs a time-series input matrix based on the selected sample data, inputs the time-series input matrix into the daily-scale gobling prediction model, and trains the daily-scale gobling prediction model using the weighted cross-entropy loss function to achieve daily-scale prediction of gobling events.

[0033] This invention provides an imbalanced sample galloping prediction device based on meteorological data, which aims to solve the problem of poor prediction performance caused by extreme sample imbalance in existing galloping prediction. By constructing a scientific sample extraction strategy, introducing a class weight mechanism, and adopting a deep learning classification model, the device improves the accuracy of galloping event identification and early warning capability while ensuring the authenticity and representativeness of the data, thereby enhancing the operational safety and intelligence level of the power grid system under complex meteorological conditions.

[0034] Thirdly, the present invention provides a computer device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the unbalanced sample jogging prediction method based on meteorological data as described in the first aspect or any corresponding embodiment.

[0035] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the unbalanced sample gobling prediction method based on meteorological data according to the first aspect or any corresponding embodiment described above. Attached Figure Description

[0036] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0037] Figure 1 This is a flowchart illustrating the unbalanced sample jogging prediction method based on meteorological data according to an embodiment of the present invention.

[0038] Figure 2 This is a structural block diagram of an unbalanced sample sway prediction device based on meteorological data according to an embodiment of the present invention;

[0039] Figure 3 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] This invention provides a method for predicting power system galloping based on imbalanced samples of meteorological data. By employing techniques such as equipment stratified sampling, category weighting, and deep learning modeling, it effectively improves the accuracy of galloping prediction under severe category imbalance conditions and enhances the power system's response capability to galloping risks.

[0042] According to an embodiment of the present invention, an embodiment of an unbalanced sample sway prediction method based on meteorological data is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0043] This embodiment provides a method for predicting imbalanced sample galloping based on meteorological data. Figure 1 This is a flowchart of an unbalanced sample galloping prediction method based on meteorological data according to an embodiment of the present invention, such as... Figure 1 As shown, the process includes the following steps:

[0044] Step S1: Install multiple monitoring devices along the transmission line to collect historical monitoring data from each monitoring point and simultaneously collect meteorological information from the corresponding monitoring points.

[0045] Specifically, multiple online monitoring devices are deployed along the transmission lines. Each device is equipped with galloping detection capabilities and transmits galloping status data in real time. Five consecutive years of daily monitoring data are collected to form a raw sample database. Simultaneously, meteorological information corresponding to the monitoring points is collected, including but not limited to: wind speed, temperature, humidity, precipitation type, and precipitation intensity. All data is indexed using a combination of "device number + timestamp" to ensure spatiotemporal consistency.

[0046] Step S2: Using the monitoring device and date as the basic unit, select one sample data from the daily data of each device, group the non-galloping samples in the sample data according to the line number, retain only the samples corresponding to different spans in each group, retain all galloping samples in the sample data, and the sample data contains the complete time series data of the day.

[0047] Specifically, to address the problem of scarce dancing samples and severe imbalance in category ratios in the monitoring data, this application adopts a stratified sampling and structural equilibrium strategy to finely screen and control the training samples.

[0048] 1) First, using the monitoring device and date as the basic unit, only one sample is retained for each monitoring device per day, and duplicate records of the same device within the same day are removed to avoid data stacking causing model deviation.

[0049] 2) Secondly, in the unmoving samples, further group them according to the line number, and retain only the samples corresponding to different spans within each line to avoid a large concentration of samples with the same span, ensuring balanced coverage of the data in the spatial structure. Let the set of all spans of a certain line be D={d1,d2,…,dn}, then only the samples corresponding to each span within the set are retained:

[0050]

[0051] 3) All dancing samples are retained without any deletions to ensure that the model can fully learn the dancing features.

[0052] This application addresses the challenges of large sample sizes and extreme class imbalance in dynamometer data by proposing a sample extraction method based on "monitoring device-date." This method selects one representative data point from each device's daily data for modeling, effectively reducing redundant data and maintaining a balanced global spatial and temporal distribution. By limiting the number of non-dynamometer samples while ensuring the complete utilization of dynamometer samples, this strategy significantly alleviates the sample imbalance problem, while improving model training efficiency and data representativeness. It is suitable for sample management and modeling needs in large-scale heterogeneous monitoring scenarios.

[0053] Step S3: Based on the characteristics of the line span, the high incidence of galloping, and typical weather combinations, assign differentiated loss weights to the sample data and construct a weighted cross-entropy loss function.

[0054] Specifically, this application innovatively introduces structural sensitivity factors, time-period adjustment factors, and meteorological condition adjustment factors into the traditional category-weighted method, constructing a multi-factor combination category weighting mechanism. This mechanism assigns differentiated loss weights to samples based on line span characteristics, high-incidence periods of galloping, and typical meteorological combinations, thereby enabling focused learning of easily missed samples and improving the model's ability to identify small galloping samples. This method exhibits good scalability and engineering adaptability, significantly enhancing the robustness and practicality of deep learning models under real-world operating conditions.

[0055] Step S4: Construct a daily-scale gobling prediction model using an LSTM network. Construct a time-series input matrix based on the selected sample data. Input the time-series input matrix into the daily-scale gobling prediction model. Train the daily-scale gobling prediction model using a weighted cross-entropy loss function to achieve daily-scale prediction of gobling events.

[0056] Specifically, to perform time-series analysis on daily meteorological data and predict whether a gobbling event will occur on a given day, this application employs a model based on a Long Short-Term Memory (LSTM) network. This application combines the sensitivity of gobbling behavior to meteorological changes, using an LSTM network to model daily meteorological time-series data to achieve daily-scale prediction of gobbling events. Compared to traditional methods that only use statistical features, this application preserves time-series patterns and abrupt change characteristics, improving the predictability and accuracy of gobbling predictions, and better meeting the practical application needs of "predicting the risk of gobbling the following day" in engineering.

[0057] In a specific example, over 3,000 online monitoring devices were deployed in a power grid area, collecting over 10 million data points over five years. Meteorological features (such as wind speed, humidity, and precipitation) were extracted. A weighted cross-entropy neural network model was used for training, increasing the galloping detection rate on the test set from 60% to over 87%. Compared with existing technologies, this invention has the following advantages: it is suitable for extremely imbalanced sample scenarios, significantly improving the minority class (galloping) detection rate; it balances accuracy and generalization ability, showing stable performance across multiple monitoring points; it is easy to deploy and expand, and can be connected to any meteorological monitoring system or online device; it enhances the power grid's safe operation and maintenance capabilities, serving as intelligent support for galloping prevention and strategy formulation.

[0058] This invention provides a method for predicting class imbalance based on meteorological data. Addressing the scarcity and spatiotemporal dispersion of class imbalance samples, it designs a sample extraction strategy based on the monitoring device-time dimension and combines class weighting factors with a deep learning model to improve the model's ability to identify minority class events. This invention fully utilizes the large-scale data resources accumulated over five years from more than 3,000 monitoring devices, effectively alleviating the class imbalance problem without introducing false samples, and significantly improving the accuracy and robustness of class imbalance prediction. It has significant engineering practical value and promising prospects for widespread application.

[0059] In one optional implementation, step S3 includes:

[0060] Step S31: Extract categorical features from meteorological information and calculate the category frequency weighting factor based on the frequency of each category in the sample data.

[0061] Specifically, the weights of the base classes are calculated based on the frequency of each class in the training set. :

[0062]

[0063] in f is the weight of category i. i The frequency of occurrence of the category is represented by c, which is an adjustment constant to prevent the denominator from being zero and to prevent extreme scaling. Generally, c is taken as 1.01.

[0064] In the dance prediction task, there are two categories: Category 0, no dance (large number of samples); Category 1, dance (very few samples).

[0065] Assume the statistics in the current sample are as follows:

[0066] Number of samples without dancing: 800,000, accounting for approximately ;

[0067] Number of dance samples: 5,000, accounting for approximately .

[0068] Let c = 1.01, then:

[0069] Category 0 (No Dancing) Weight:

[0070] Category 1 (Dancing) Weight:

[0071]

[0072] In this way, the loss of the dancing samples is multiplied by a larger w1, making the model pay more attention to this rare data, thereby improving the recognition rate of dancing and avoiding being ignored due to the small number of samples.

[0073] Step S32: Based on the differences in the risk of galloping at different gear intervals, set corresponding structural sensitivity factors for different gear intervals.

[0074] Specifically, considering the differences in the risk of galloping at different gear intervals, a structural factor is introduced. Distinguishing weights are assigned to samples that do not move:

[0075]

[0076] in, The structural factor corresponding to the j-th interval is set based on historical dance statistics or experience. Common long intervals are typically set as follows: >1, Normal gear distance set to =1.

[0077] Step S33: Based on the probability of the dance occurring in different time periods, set corresponding time period sensitivity factors for different time periods.

[0078] Specifically, considering that the peak times for dancing (such as nighttime and early morning) are easily overlooked by the model, a time-weighted term is specially set up. :

[0079]

[0080] in, The sensitivity coefficient for the t-th time period, such as nighttime. =1.5, daytime =1.

[0081] Step S34: Set meteorological condition adjustment factors for sample data that meet the preset meteorological combination.

[0082] Specifically, for samples under specific weather combinations (such as high wind speed and high humidity), the samples are further multiplied by a weather adjustment factor. :

[0083]

[0084] If the sample meets specific meteorological conditions (e.g., wind speed > 12 m / s, humidity > 90%), then a setting is established. >1, otherwise 1.

[0085] In one alternative implementation, the method further includes:

[0086] Step S5: Normalize the numerical features in the meteorological information, including wind speed, temperature, humidity and precipitation.

[0087] Step S6: Convert the categorical features in the meteorological information into uniquely coded features. The categorical features include wind direction category and precipitation category.

[0088] Specifically, meteorological factors influencing power line galloping are extracted as input features for the model, primarily including: wind speed (m / s), wind direction, temperature (°C), humidity (%), precipitation (mm), and precipitation type (coded). All numerical features, such as wind speed, temperature, humidity, and precipitation, are normalized. Categorical features, such as wind direction and precipitation type, are converted into one-heat coded features. This allows the model to more clearly distinguish different wind directions and precipitation types, thus more accurately learning the relationship between these features and power line galloping.

[0089] In one optional implementation, to perform time-series analysis on daily meteorological data and predict whether a dancing event will occur on that day, this application employs a model based on a Long Short-Term Memory (LSTM) network. This model serves only as a tool for implementing the business, focusing on improving prediction performance in conjunction with the aforementioned multi-factor loss weighting mechanism. Its basic structure and training process are described below:

[0090] 1) Input data construction

[0091] For each monitoring device, a time-series input matrix containing 24 hours of meteorological data is constructed daily, with the corresponding label y∈{0,1} indicating whether a dancing event occurred that day.

[0092] 2) Model Structure

[0093] The model uses one or two LSTM layers to process the temporal input matrix, extracts temporal features, maps them to a single output node through a fully connected layer, and then outputs the probability of the dance occurring on that day through a Sigmoid activation function. The specific expression is as follows:

[0094]

[0095] in, Use the Sigmoid activation function; This is the weight matrix of the fully connected layer, which is usually 1×d, where d is the dimension of the hidden state vector output by the LSTM (e.g., 64 or 128). This represents the hidden state of the LSTM network at the last time step. The bias term is a scalar.

[0096] 3) Model training and loss function

[0097] This application employs a weighted cross-entropy loss function constructed using the aforementioned multi-factor weighting mechanism to train the model. The loss function is:

[0098]

[0099] in, The weighted cross-entropy loss function is... The final sample weights for multi-factor synthesis. represents the true label (0 or 1) of the i,j,t,k sample. This represents the probability of a gobling event predicted by a daily-scale gobling prediction model.

[0100] The model was trained using a conventional optimizer (such as Adam) and an early stopping strategy. During training, various metrics of the model on the validation set were recorded to ensure that the model could achieve high recall and F1 score even under imbalanced sample conditions.

[0101] This embodiment also provides an imbalanced sample gyratory prediction device based on meteorological data. This device is used to implement the above embodiments and preferred embodiments, and details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0102] This embodiment provides a device for predicting imbalanced sample galloping based on meteorological data, such as... Figure 2 As shown, it includes:

[0103] The data acquisition module 21 is used to deploy multiple monitoring devices along the transmission line, collect historical monitoring data from each monitoring point, and simultaneously collect meteorological information from the corresponding monitoring points.

[0104] The data filtering module 22 is used to select one sample data from the daily data of each device, based on the monitoring device and date as the basic unit. The non-vibrating samples in the sample data are grouped by line number, and only samples corresponding to different ranges are retained in each group. All vibrating samples in the sample data are retained. The sample data includes the complete time series data of the day.

[0105] The weight assignment module 23 is used to assign differentiated loss weights to sample data based on line span characteristics, high-incidence periods of galloping, and typical weather combinations, and to construct a weighted cross-entropy loss function.

[0106] The model prediction module 24 is used to construct a daily-scale gobling prediction model using an LSTM network. It constructs a time-series input matrix based on the selected sample data, inputs the time-series input matrix into the daily-scale gobling prediction model, and trains the daily-scale gobling prediction model using a weighted cross-entropy loss function to achieve daily-scale prediction of gobling events.

[0107] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0108] In this embodiment, the unbalanced sample gobble prediction device based on meteorological data is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0109] This invention provides an imbalanced sample galloping prediction device based on meteorological data, which aims to solve the problem of poor prediction performance caused by extreme sample imbalance in existing galloping prediction. By constructing a scientific sample extraction strategy, introducing a class weight mechanism, and adopting a deep learning classification model, the device improves the accuracy of galloping event identification and early warning capability while ensuring the authenticity and representativeness of the data, thereby enhancing the operational safety and intelligence level of the power grid system under complex meteorological conditions.

[0110] This invention also provides a computer device having the above-described features. Figure 2 The device shown is an imbalanced sample gobble prediction device based on meteorological data.

[0111] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of a computer device provided in an optional embodiment of the present invention, such as... Figure 3 As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 3 Take a processor 10 as an example.

[0112] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0113] The memory 20 stores instructions executable by at least one processor 10 to cause the at least one processor 10 to perform the method shown in the above embodiments.

[0114] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0115] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0116] The computer device also includes a communication interface 30 for communicating with other devices or communication networks.

[0117] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0118] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for predicting imbalanced sample galloping based on meteorological data, characterized in that, The method includes: Multiple monitoring devices are deployed along the transmission line to collect historical monitoring data from each monitoring point and simultaneously collect meteorological information from the corresponding monitoring points. Using monitoring devices and dates as basic units, a sample data point is selected from the daily data of each device. The non-galloping samples in the sample data are grouped by line number. Only samples corresponding to different ranges are retained in each group. All galloping samples in the sample data are retained. The sample data contains complete time-series data for the day. Based on the characteristics of the line span, the high-incidence period of galloping and typical weather combinations, the sample data are assigned differentiated loss weights, and a weighted cross-entropy loss function is constructed. An LSTM network is used to construct a daily-scale gobling prediction model. A time-series input matrix is ​​constructed based on the selected sample data. The time-series input matrix is ​​input into the daily-scale gobling prediction model. The weighted cross-entropy loss function is used to train the daily-scale gobling prediction model to achieve daily-scale prediction of gobling events. Based on the characteristics of the line span, the peak periods of galloping, and typical weather combinations, differential loss weights are assigned to the sample data, including: Extract categorical features from the meteorological information, and calculate the category frequency weighting factor based on the frequency of each category in the sample data; Based on the differences in the risk of galloping at different gear intervals, corresponding structural sensitivity factors are set for different gear intervals; Based on the probability of the dance occurring in different time periods, corresponding time-sensitivity factors are set for different time periods; For sample data that meet the preset meteorological combination, set meteorological condition adjustment factors.

2. The method for predicting imbalanced sample agitation based on meteorological data according to claim 1, characterized in that, The weighted cross-entropy loss function is as follows: in, The weighted cross-entropy loss function is... The final sample weights for multi-factor synthesis. For the true labels of the i,j,t,k samples, This represents the probability of a gobling event predicted by a daily-scale gobling prediction model.

3. The method for predicting imbalanced sample agitation based on meteorological data according to claim 1, characterized in that, The method further includes: The numerical features in the meteorological information are normalized, including wind speed, temperature, humidity, and precipitation. The categorical features in the meteorological information are transformed into uniquely coded features, which include wind direction category and precipitation category.

4. The method for predicting imbalanced sample agitation based on meteorological data according to claim 1, characterized in that, All sample data are indexed using a combination of device number and timestamp.

5. The method for predicting imbalanced sample galloping based on meteorological data according to claim 1, characterized in that, The daily-scale gobling prediction model uses a one- or two-layer LSTM network to process the time-series input matrix, extracts time features, maps them to a single output node through a fully connected layer, and then outputs the probability of the gobling occurring on that day through a Sigmoid activation function.

6. The method for predicting imbalanced sample galloping based on meteorological data according to claim 5, characterized in that, The diurnal-scale gobling prediction model is expressed as follows: in, It is the Sigmoid activation function. This is the weight matrix of the fully connected layer. This represents the hidden state of the LSTM network at the last time step. This is a bias term.

7. A device for predicting imbalanced sample agitation based on meteorological data, characterized in that, The device includes: The data acquisition module is used to deploy multiple monitoring devices along the transmission line to collect historical monitoring data from each monitoring point and simultaneously collect meteorological information from the corresponding monitoring points. The data filtering module is used to select one sample data from the daily data of each device, using the monitoring device and date as the basic unit. The non-vibrating samples in the sample data are grouped by line number, and only samples corresponding to different ranges are retained in each group. All vibrating samples in the sample data are retained. The sample data contains complete time-series data for the day. The weight assignment module is used to assign differentiated loss weights to sample data based on line span characteristics, high-incidence periods of galloping, and typical weather combinations, and to construct a weighted cross-entropy loss function. The model prediction module is used to construct a daily-scale gobling prediction model using an LSTM network, construct a time-series input matrix based on the selected sample data, input the time-series input matrix into the daily-scale gobling prediction model, and train the daily-scale gobling prediction model using the weighted cross-entropy loss function to achieve daily-scale prediction of gobling events. Based on the characteristics of the line span, the peak periods of galloping, and typical weather combinations, differential loss weights are assigned to the sample data, including: Extract categorical features from the meteorological information, and calculate the category frequency weighting factor based on the frequency of each category in the sample data; Based on the differences in the risk of galloping at different gear intervals, corresponding structural sensitivity factors are set for different gear intervals; Based on the probability of the dance occurring in different time periods, corresponding time-sensitivity factors are set for different time periods; For sample data that meet the preset meteorological combination, set meteorological condition adjustment factors.

8. A computer device, characterized in that, include: The system includes a memory and a processor, which are interconnected. The memory stores computer instructions, and the processor executes the computer instructions to perform the unbalanced sample jogging prediction method based on meteorological data as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the unbalanced sample gobling prediction method based on meteorological data as described in any one of claims 1 to 6.

Images